| AlexDaniel | timotimo: hmm… what's the difference? | 00:04 | |
| “Quiet" mode is pretty much like normal mode, in that it will still register your activity, and notice that you need to take a break, but it will not actually prompt you to take one, nor block you from using the computer” | 00:05 | ||
| timotimo | yeah | 00:07 | |
| how does it get in your way? | |||
|
00:14
Kaiepi joined
|
|||
| AlexDaniel | timotimo: it breaks the flow | 00:36 | |
| timotimo | oh | ||
| can always postpose | |||
| postpone* | |||
| AlexDaniel | yeah or just disable :) | 00:37 | |
| at first it was ok, then I started noticing that I am postponing way too much… then I gave up :) | 00:38 | ||
| it's ok though, as it seems that my RSI is not coming back | 00:39 | ||
| MasterDuke | inclusive >100% seems like it could be fine | ||
| AlexDaniel | and my eyes, hmm… I think I just need glasses | ||
| timotimo | there's no shame in that | 00:47 | |
| MasterDuke | i am quite blind without them, worn them every day of my life since i was young | 00:51 | |
| Geth | MoarVM/profiler_new_spesh_semantics: e14e666cf2 | (Timo Paulssen)++ | src/profiler/instrument.c let sequence numbers start at 0 in the profile i don't think it'd be worth much to have them match up exactly with the internal numbers (in which case I'd only subtract 1) |
00:52 | |
| MasterDuke | does the "inclusive can go >100%" question have to be answered in moarvm? or can that be a toggle in the profile viewer? | 00:59 | |
| timotimo: ^^^ | 01:04 | ||
| timotimo | the profile viewer would do it | 01:11 | |
| MasterDuke | so you could just have a "scaled view" option or something like that? | 01:18 | |
| samcv | AlexDaniel: i'll get the release ready later tonight | ||
| since it seems all is ok? | |||
| at least from what i read in #perl6-dev | |||
| timotimo | that'd work | 01:20 | |
| MasterDuke | timotimo: oh, btw, both the close button on the bottom right and the x in the upper right don't work for me (FF 58 on kubuntu 17.10) on pop-up windows in the profile (e.g., after clicking "view" in the "allocating routines" column) | 01:23 | |
| timotimo | yeah, known for years ;) | 01:27 | |
| MasterDuke | heh | 01:28 | |
| AlexDaniel | samcv: yeah, everything is ok | 01:40 | |
|
02:57
ilbot3 joined
03:29
greppable6 joined
07:40
brrt joined
|
|||
| brrt | good * #moarvm | 07:40 | |
| yoleaux | 18 Feb 2018 15:30Z <nine> brrt: MVM_SPESH_BLOCKING=1 MVM_SPESH_NODELAY=1 perl6 -Ilib t/spec/S09-typed-arrays/hashes.t | ||
| 18 Feb 2018 15:30Z <nine> brrt: fails with "Incompatible MROs in P6opaque rebless for types NQPClassHOW and QAST::WVal+{QAST::SpecialArg}" but passes with MVM_JIT_EXPR_DISABLE=1 | |||
| 18 Feb 2018 15:59Z <nine> brrt: Please disregard that. I didn't realize that I was running with an experimental patch applied. | |||
| brrt | ohai nine | ||
| i almost had something similar pinned down, but this morning it's 'gone' | 07:41 | ||
|
07:51
domidumont joined
07:57
domidumont joined
08:13
brrt joined
|
|||
| Geth | MoarVM/jit-expr-optimizer: 13 commits pushed by (Bart Wiegmans)++ review: github.com/MoarVM/MoarVM/compare/a...df8272c7c2 |
08:40 | |
|
09:18
zakharyas joined
09:21
zakharyas joined
09:27
zakharyas joined
09:42
zakharyas joined
11:28
zakharyas joined
11:46
Kaiepi joined
12:14
zakharyas joined
12:52
dogbert2 joined
|
|||
| MasterDuke | brrt: i don't remember where the debugging left off. you figured out the getlexstatic_o problem? | 13:39 | |
| brrt | did not | 13:41 | |
| MasterDuke | ah. anything more i could get out of lldb to help? | 13:42 | |
| brrt | ehm, maybe, but it's a pretty slow interface this way :-) | ||
| MasterDuke | heh, that it is | ||
|
13:58
domidumont1 joined
14:06
zakharyas joined
|
|||
| MasterDuke | brrt: which of your recent commits was a fix for box_i? | 14:13 | |
| hm, maybe it wasn't specifically for box_i, but fixed it also | 14:15 | ||
|
14:18
zakharyas joined
|
|||
| brrt | i think the if_o -> istrue + if_o split | 14:21 | |
| MasterDuke | k, thanks | 14:22 | |
| Geth | MoarVM/jit-expr-optimizer: 95cffa2567 | (Bart Wiegmans)++ | 2 files Restore ^func macro Even though it is now useless, it doesn't hurt either, and it keeps us synchronized with master. |
14:23 | |
|
14:35
domidumont joined
|
|||
| MasterDuke | brrt: how do i pass a NULL to a function in a template? | 14:37 | |
| brrt | just (const 0) | 14:38 | |
| (const 0 ptr_sz) | |||
| MasterDuke | k, thanks | ||
| brrt | :-) | ||
|
14:41
zakharyas joined
14:48
zakharyas joined
|
|||
| MasterDuke | huh. box_i is fine, but adding box_s causes `Cannot look up attributes in a VMNull type object at src/Perl6/Grammar.nqp:4018` when building rakudo | 14:58 | |
|
15:42
zakharyas joined
16:00
releasable6 joined
16:01
brrt joined
16:19
buggable joined
16:20
buggable joined,
huggable joined
|
|||
| MasterDuke | that's being thrown from getattr_s, `if (!IS_CONCRETE(obj)) MVM_exception_throw_adhoc(tc, "Cannot look up attributes in a %s type object", MVM_6model_get_debug_name(tc, obj));` | 16:21 | |
|
16:21
ZofBot joined
|
|||
| MasterDuke | .ask brrt is this something i could use the jit-bisect tool for? irclog.perlgeek.de/moarvm/2018-02-19#i_15834025 | 16:32 | |
| yoleaux | MasterDuke: I'll pass your message to brrt. | ||
|
16:33
brrt joined
16:52
nativecallable6 joined,
bisectable6 joined,
releasable6 joined,
bloatable6 joined,
reportable6 joined,
committable6 joined,
quotable6 joined,
coverable6 joined,
notable6 joined,
benchable6 joined
16:53
evalable6 joined,
statisfiable6 joined,
unicodable6 joined,
greppable6 joined,
squashable6 joined,
shareable6 joined
|
|||
| MasterDuke | brrt: i just tried running `perl ..//MoarVM/tools/jit-bisect.pl -- <the command to compile rakudo's core setting>`. will that do what i want? | 16:59 | |
| brrt | usually yes | ||
| yoleaux | 16:32Z <MasterDuke> brrt: is this something i could use the jit-bisect tool for? irclog.perlgeek.de/moarvm/2018-02-19#i_15834025 | ||
| MasterDuke | the rakudo compile doesn't segfault, just exits from a thrown exception | 17:00 | |
| brrt | doesn't matter, as long as it exits anomalously | ||
| MasterDuke | cool | ||
| working on last frame = 128 now | 17:01 | ||
| probably should have run it on my desktop | |||
| ah, 256 was not ok | 17:02 | ||
|
17:13
bisectable6 joined
|
|||
| MasterDuke | .ask brrt there wasn't any output from the two jit-comparify-asm.pl calls at the end, is that expected? | 17:26 | |
| yoleaux | MasterDuke: I'll pass your message to brrt. | ||
| MasterDuke | oh, there are some log files created | ||
| .ask brrt if i put the two sets of .bin, .asm, and .log files somewhere you could get them would you be able to use them? | 17:28 | ||
| yoleaux | MasterDuke: I'll pass your message to brrt. | ||
|
17:42
domidumont joined
18:16
bart__ joined
|
|||
| brrt | . | 18:16 | |
| yoleaux | 17:26Z <MasterDuke> brrt: there wasn't any output from the two jit-comparify-asm.pl calls at the end, is that expected? | ||
| 17:28Z <MasterDuke> brrt: if i put the two sets of .bin, .asm, and .log files somewhere you could get them would you be able to use them? | |||
| brrt | yes, and yes, on both counts | ||
| MasterDuke | cool, let me put them somewhere... | 18:19 | |
| brrt | you can skip the .bins, they might not pass the malwarwe filters | 18:20 | |
| MasterDuke | try gist.github.com/MasterDuke17/be6b6...3d39d0215a | 18:21 | |
|
18:34
domidumont joined
20:18
brrt joined
|
|||
| samcv | brrt: any clue how much speed increase we get when an op has a JIT expression template added? | 20:23 | |
| brrt | per op, negligible | 20:24 | |
| or 'it depends' | |||
| MasterDuke | the value comes from being able to combine multiple ops/blocks and perform optimizations on the combination, so a missing template forces an end to optimization potential | 20:32 | |
| is my understanding | |||
| brrt: is ^^^ somewhat correct? | 20:33 | ||
| samcv | well. that description sounds like A Good Thing™ | 20:34 | |
| brrt | that is more or less correct | 20:39 | |
| and in some cases, the expr jit can generate better code | 20:40 | ||
| but the difficulty is that it effectively competes against handwritten assembly | |||
| samcv | competes? in what sennse | ||
| *sense | |||
| MasterDuke | compared to the "lego" jit i believe | 20:43 | |
| brrt | yeah | ||
| the lego jit composes together handwritten pieces | 20:44 | ||
| the expr jit compiles together invidiual instructions | |||
| MasterDuke | brrt: btw, see anything in those log files? | 20:45 | |
| brrt | people who write assembly by hand can be cleverer than programs | ||
| probably, but i'm not actively looking atm :-) | |||
| i'm trying to debug my washing machine | |||
| well, i've debugged it, with the help of that wonderful thing called The Internet | |||
| MasterDuke | heh. doubt that involves looking at assembly | 20:46 | |
| brrt | it does involve disassembly though | 20:52 | |
| MasterDuke | of drum memory? | 20:55 | |
| brrt | not the drum, the water inlet valve, as a matter of fact | 20:59 | |
| timotimo | water has memory! | 22:46 | |
| (wow, i'm fashionably late to this party) | |||
| samcv | Disable getlexstatic_o for the time being # is this still disabled? | 23:01 | |
| MasterDuke | samcv: yeah | 23:20 | |
| samcv | ok cool | ||
| lizmat | And another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2018/02/19/...6ya-giris/ | 23:29 | |
| timotimo | yay | 23:30 | |